#30 Panics an OSX box

[Anonymous]

Got a couple of panics recently on my Mac when trying to mount a
Lacie drive (connected via IEEE1394) with a single ext3 partition. It
happened twice, the first time with an older version of your code,
the second with the latest 1.3. Here's the full story:

1. Drive had a perfectly good ext2 partition created on the Mac,
which I could read/write to, but when I plugged that in to the Linux
box, I got "bad magic number", so couldn't mount it.

2. Created a new ext2 partition on the linux box, moved it back to
the mac, plugged it in. Panic, machine reboots, plugged the drive
in again and it started to fsck and didn't mount.

3. Downloaded and installed 1.3 version of ext2fs for OSX (not
sure what previous version was).

4. Created a new ext3 partition on the drive using "mke2fs -j /dev/
sdc1" from my Linux box, which is running stock Redhat 9.
Mounted the drive, copied a 4.3Gb file across, unmounted it, all
without drama

5. Plugged the drive into my mac (running 10.3.5 and your code, v
1.3), got another panic.

I'm attaching the logs - these include the crash from the earlier
version, which may or may not be of any use to you. There's some
note about misalignment in there, but as far as I can tell the drive
is fine - they're designed to be portable after all, and, certainly I
can read/write to it from Linux no problem, and the drive itself is
lying flat on my desk when I plug it in.

I'm mike at big dot faceless dot org if you need to get in contact.

[Nobody/Anonymous]

System log and panic log

[Nobody/Anonymous]

Logged In: NO

Just a quick followup - it's the drive I think. I can mount another identical
drive without any issues, and am now unable to mount the original disk
under linux. Guess those portable Lacies aren't so portable after all.

[Brian Bergstrand]

Logged In: YES
user_id=595265

Well, I think you're right. From the logs, the problem appears hardware
related.

It looks like the drive is disappearing from the FW bus. There are quite a
few of the following entries in your system log:

Aug 31 12:37:31 localhost kernel: disk2s1: media is not pr media is not
present.
Aug 31 12:37:31 localhost kernel: disk2s1: media is not present.
Aug 31 12:37:31 localhost last message repeated 66 times
...
Aug 31 12:38:28 localhost kernel: disk3s1: alignment error.
Aug 31 12:38:58 localhost last message repeated 299 times

Looking at the panic log seems to confirms this:

panic(cpu 0): mount: lost mount

0x1eda4 <panic+472>: lwz r0,-32212(r29)
0xc53e4 <mount+2380>: lis r2,52
0xc510c <mount+1652>: addi r0,r26,13008

Looking at the darwin code in question, mount() calls ext2 to mount the
volume, then calls checkdirs() to move any vnodes from the old mount
to the new mount. checkdirs() calls the ext2 VFS_ROOT() op which
fails (most likely because the hardware device is now gone) and it
returns an error. This causes checkdirs() to panic.

I'd say checkdirs() shouldn't panic because of an error return, but if it
didn't you'd probably run into problems at a later point because there
would be some dangling vnodes.

I'll leave this bug open for the next week just in case you come up with
any info pointing to a driver problem.